Social Media and Elections - State of Surveillance

TL;DR: Global democracy has fallen to 1985 levels. 5.7 billion people now live under autocracy. 87% of internet users are concerned about disinformation affecting elections. Deepfake content increased 5x in recent years, with projections showing 8 million deepfakes by 2025. 68% say social media is where disinformation is most widespread. Bot networks amplify division while algorithms reward polarization. Average daily social media use exceeds 2 hours: 2 hours of algorithmically curated political reality. Researchers call this an "unprecedented threat to democracy." The platforms profit from division. Democracy requires consensus. These are incompatible.

The Stakes

This isn't about inconvenience. It's about whether democracy functions [1].

Global democratic decline (V-Dem Institute):

Democracy levels have fallen to 1985
5.7 billion people now live under autocracy
That number increased after 2024 elections
The trend is accelerating

The concern levels (Ipsos):

87% are concerned about disinformation affecting elections
47% are "very concerned"
68% say social media is where disinformation is most widespread

The exposure:

5.64 billion internet users worldwide (68.7% of population)
5.31 billion social media accounts
143-147 minutes average daily use
This level of saturation creates a "fertile environment for disinformation to propagate rapidly"

When researchers describe the current moment as an "unprecedented threat to democracy," they mean it literally. Nothing like this has existed before.

The Disinformation Machine

Election disinformation has become industrialized [2].

The actors:

Foreign governments: Russia, China, Iran running influence operations
Domestic operatives: political consultants using dark arts
Ideological networks: true believers spreading what they want to be true
Profit-motivated creators: fake news for ad revenue
AI systems: generating content at unprecedented scale

The techniques:

Fake accounts posing as concerned citizens
Coordinated posting to create illusion of consensus
Amplification of authentic extremist content
Targeting voters in swing districts with tailored lies
Creating fake "local news" sites
Deepfakes of candidates saying things they never said

Why it works:

Platforms reward engagement. Outrage engages. False, emotionally charged content spreads faster than accurate, nuanced information. The system isn't failing: it's working exactly as designed, and that design is incompatible with informed democracy.

Deepfakes: The Escalating Threat

AI-generated political content is exploding [3].

The growth:

5x local increase in deepfake content
550% global increase in known deepfakes since 2019
500,000 deepfakes shared on social media in 2023
Projections: up to 8 million by 2025

Political applications:

Candidates appearing to make statements they never made
Fabricated endorsements from celebrities or officials
Fake scandal videos timed for maximum damage
Cloned voices for robocalls and audio messages

The detection problem:

Research shows only 0.1% of people can correctly identify all deepfakes. Human detection accuracy for high-quality deepfake video is 24.5%. By the time a fake is debunked, it's already influenced millions.

The 2024 escalation:

Research found that AI influence on the 2024 election increased significantly compared to 2016, and influence is likely to increase in coming years. Social media platforms increased their role in spreading deepfakes in 2024 compared to 2016.

Bot Networks: Manufactured Consensus

Automated accounts create the illusion that fringe positions are mainstream [4].

What bots do:

Amplify content to make it appear popular
Create artificial trends and hashtags
Drown out authentic voices with coordinated noise
Make extremist positions seem like common sense
Harass and silence opposition voices

Research approaches:

Studies use machine learning models like Random Forest, XGBoost, and Botometer to detect bot activity across Facebook, Reddit, and X. Results show significant implications for "online discourse, polarization, and information dissemination leading up to elections."

The perception problem:

When you see a position supported by thousands of accounts, you assume it represents real people. Bot networks exploit this. They manufacture the appearance of consensus where none exists, shifting perception of what's normal or popular.

PLOS One research on X:

Musk's takeover showed "no reduction in inauthentic activity." The bots remain while content moderation has collapsed.

Algorithmic Polarization

The algorithms don't show you reality. They show you what engages you, which is often what enrages you [5].

Research findings (Science, 2024):

"The systematic amplification of a specific category of content (politically hostile, antidemocratic, emotionally charged) nudges users toward higher affective polarization."

How it happens:

Algorithms learn what keeps you scrolling
Hostile content about "the other side" engages
You see more content showing opponents as evil/stupid
Your perception of the other side distorts
Compromise becomes unthinkable because they're monsters

The anger amplification:

Research shows "social media can lead to much more polarized environments, where there's more anger and hostility between different groups and individuals with different political ideologies."

The democracy problem:

Democracy requires enough shared reality to negotiate. When platforms profit by showing each side a version of the other designed to maximize outrage, shared reality becomes impossible.

Microtargeting: Personalized Manipulation

Political advertising has become precision manipulation [6].

How microtargeting works:

Platforms collect detailed behavioral profiles
Campaigns buy access to highly specific audiences
Different voters see different messages
Messages can be contradictory: nobody sees both
Promises and attacks are tailored to individual psychology

The accountability problem:

Traditional TV ads are seen by everyone, visible to fact-checkers
Microtargeted ads are seen only by their targets
Lies can be precisely directed to those who'll believe them
Accountability through public scrutiny becomes impossible

Psychographic targeting:

Cambridge Analytica pioneered using psychological profiles to target voters with messages designed to exploit their specific fears and motivations. The techniques didn't disappear when the company did. They're now standard practice.

From Online to Violence

Social media disinformation leads to real-world political violence [7].

The pattern:

False narratives spread online
Communities form around shared grievance
Echo chambers intensify beliefs
Violence becomes normalized in rhetoric
Some individuals act on what they've been told

Documented incidents:

January 6, 2021: US Capitol attack organized on social media
2023: Brazilian Congress attack by election deniers coordinated online
Ongoing: Threats to election workers traced to online disinformation

Research conclusion:

"Social media disinformation, online political polarization and the use of social media to organize real-world violence harms democracy."

The platforms aren't just distorting political reality. They're enabling the organization of violence against democratic institutions.

Platform Failure

The platforms have proven unable or unwilling to address the problem [8].

Structural conflicts:

Engagement = revenue
Outrage = engagement
Disinformation often generates outrage
Limiting disinformation limits revenue
Shareholders > democracy

Moderation failures:

Content moderation teams cut for cost savings
AI moderation easily evaded
Non-English content barely moderated at all
Speed of viral spread exceeds speed of fact-checking

Policy reversals:

X under Musk: reduced content moderation, reinstated banned accounts
Meta: announced end to third-party fact-checking in 2025
Platform policies change based on political pressure, not effectiveness

Expecting platforms to protect democracy against their financial interests has proven naive. Voluntary self-regulation has failed.

Regulatory Responses

Governments are attempting to respond, with mixed results [9].

Legislative efforts:

Proposals mandating disclosure of AI-generated political content
Meta now requires advertisers to label AI-altered content in political ads
EU Digital Services Act imposes transparency requirements
Various state laws on election-related deepfakes

Proposed solutions:

Congressional action on platform transparency
Accessible voter education
Mandated disclosure of ad targeting criteria
Algorithmic accountability requirements

The enforcement problem:

Platforms operate globally; laws are national
Content spreads faster than enforcement
Regulation lags technology by years
Political will varies with who benefits from disinformation

What You Can Do

Individual action has limits but isn't useless [10].

Information hygiene:

Verify before sharing, especially if content angers you
Check sources outside social media
Be skeptical of content that perfectly confirms your beliefs
Use fact-checking resources for claims about candidates

Reduce algorithmic exposure:

Use chronological feeds when possible
Follow trusted sources directly rather than relying on algorithm
Limit time on platforms during election seasons
Get news from multiple sources with different perspectives

Community resistance:

Correct misinformation when you see it in your networks
Support local journalism
Help less tech-savvy voters understand manipulation tactics

Political engagement:

Support platform accountability legislation
Advocate for algorithmic transparency
Vote for candidates who understand these threats
Support election security measures

The Bottom Line

Democracy has declined globally to 1985 levels. 5.7 billion people live under autocracy. And every day, 5.3 billion social media users spend over two hours consuming algorithmically curated content designed to engage through outrage, amplified by bot networks, contaminated with disinformation, and increasingly featuring AI-generated content indistinguishable from reality.

This isn't a coincidence. The platforms profit from engagement. Division engages. They've built systems that reward the most outrageous, polarizing, enraging content because that content keeps people scrolling. When that content is political disinformation, the platforms profit just the same.

Democracy requires shared reality. It requires enough common ground to negotiate differences. It requires trust in institutions and information sources. Social media systematically undermines all of these because doing so is profitable.

The platforms have proven unable to self-regulate. Voluntary measures have failed. Content moderation is being abandoned. AI is making fabrication easier while detection lags behind. Every election cycle the tools for manipulation improve while the defenses weaken.

Researchers call this an "unprecedented threat." They're not exaggerating. Nothing in human history has allowed this level of precision manipulation at this scale. We're running a global experiment with democracy as the test subject and engagement metrics as the only success criterion.

The platforms won't save democracy. They're not designed to. The only forces that can are informed citizens who understand the manipulation, and regulations that change the incentive structure. Until those forces overcome the platforms' interest in division, democracy will continue to decline.

The question isn't whether social media affects elections. The evidence is overwhelming. The question is whether we do anything about it before the damage is irreversible.